Work

Decision letter: Evolution of natural lifespan variation and molecular strategies of extended lifespan in yeast

Year: 2021
Type: peer-review
Abstract: Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract To understand the genetic basis and selective forces acting on longevity, it is useful to examine lifespan variation among closely related species, or ecologically diverse isolates of the same species, within a controlled environment. In particular, this approach may lead to understanding mechanisms underlying natural variation in lifespan. Here, we analyzed 76 ecologically diverse wild yeast isolates and discovered a wide diversity of replicative lifespan (RLS). Phylogenetic analyses pointed to genes and environmental factors that strongly interact to modulate the observed aging patterns. We then identified genetic networks causally associated with natural variation in RLS across wild yeast isolates, as well as genes, metabolites, and pathways, many of which have never been associated with yeast lifespan in laboratory settings. In addition, a combined analysis of lifespan-associated metabolic and transcriptomic changes revealed unique adaptations to interconnected amino acid biosynthesis, glutamate metabolism, and mitochondrial function in long-lived strains. Overall, our multiomic and lifespan analyses across diverse isolates of the same species shows how gene–environment interactions shape cellular processes involved in phenotypic variation such as lifespan. Editor's evaluation This manuscript characterizes differences in replicative lifespan across a collection of natural yeast isolates. Additionally, it provides transcriptional and metabolomic resources that help shed important new light on the mechanistic bases of functional differences among the wild yeast isolates. The revised version further tests the impact on lifespan in short- and long-lived strains of target genes identified. This work has important implications for the fields biology of aging and potentially evolutionary biology. https://doi.org/10.7554/eLife.64860.sa0 Decision letter eLife's review process Introduction Diverse selective forces (mutation, selection, and drift) generate enormous variation within and among species (Via and Lande, 1985; Li et al., 2018). Consequently, many morphological, behavioral, and physiological phenotypes (traits) vary within and between species in natural populations (Wagner and Zhang, 2011; Sambandan et al., 2008; Malagón et al., 2014; Caswell, 2007; Fusco and Minelli, 2010). For example, genetic variation in natural populations of many organisms can differentially affect their neural and endocrine functions, leading to variation in quantitative life-history traits such as fitness and age at maturation (Bonier et al., 2009; Finch and Rose, 1995). Variation in another fitness trait, lifespan, has also attracted much attention (Arking et al., 1996; Libert and Pletcher, 2007; Ratikainen and Kokko, 2019). Across eukaryotic species, longevity can differ over many orders of magnitude, from days to centuries, for example the Greenland shark (Somniosus microcephalus) may live for more than 500 years whereas some species live only several days (Mortimer and Johnston, 1959; Nielsen et al., 2016; Jones et al., 2014). Longevity also varies among individuals of the same species, indicating that variability of lifespan is not constrained at the level of species, and that the molecular determinants of lifespan vary within the same genetic pool (Finch and Pike, 1996; Klass, 1983; Kaya et al., 2015; Ma et al., 2018; Dato et al., 2018; Wright et al., 2019). What, then, are the factors that determine the lifespan of individuals? The molecular pathways that underlie the genetic and environmental determinants of lifespan are among the most intensely studied areas in the aging field (Kaya et al., 2015; Ma et al., 2018; de Magalhães et al., 2012; Li and de Magalhães, 2013; Ma and Gladyshev, 2017; van Dongen et al., 2016; Dato et al., 2016). Laboratory animal models have shown that longevity can be extended by environmental (Dillon et al., 2007; Johnston and Snell, 2016; Packer and Fuehr, 1977; Bisschops et al., 2015), dietary (Alic and Partridge, 2011; Stefana et al., 2017; Kitada et al., 2019), pharmacological (Bitto et al., 2016; Novelle et al., 2016; Rajman et al., 2018; Nadon et al., 2017), and genetic interventions (Zwaan, 1999; Vijg and Suh, 2005; Singh et al., 2019). However, many of these laboratory-adapted populations are constrained by genetic and environmental background (Harper, 2008; de Magalhães, 2014; Reichard, 2016). For example, artificially created mutant strains may show longer lifespan under laboratory settings but demonstrate reduced fitness in their natural environment (Bartke et al., 2001; Jenkins et al., 2004; Chen et al., 2007; Delaney et al., 2011). A more integrated approach is needed to understand how the natural environment and natural selection interact to shape lifespan and associated life-history traits. In order to better understand the impact of natural genetic variation on lifespan, we studied 76 wild isolates of the budding yeast to capture the molecular signatures of evolved diversity of lifespan. This collection included 40 diploid isolates of Saccharomyces cerevisiae and 36 diploid isolates of Saccharomyces paradoxus (Hyma and Fay, 2013; Liti et al., 2009). Their niches include human-associated environments, such as breweries and bakeries, and different types of wild ecological niches, such as trees, fruits, vineyards, and soils across different continents. There was also a group of clinical isolates; S. cerevisiae strains isolated from immunocompromised patients. S. cerevisiae and S. paradoxus are closely related (share a common ancestor between 0.4 and 3 million years ago), with 90% genome identity, and can mate and produce viable progeny (Liti et al., 2006; Tirosh et al., 2009). Earlier genome sequencing of these isolates revealed allelic profiles, their ploidy status, and a phylogeny of these isolates (Hyma and Fay, 2013; Liti et al., 2009; Liti et al., 2006). Based on these studies, it was shown that while some of these strains fall into distinct lineages with unique genetic variants, almost half of the strains have mosaic recombinant genomes arising from outcrosses between genetically distinct lineages of the same species (Liti et al., 2009). Because of their wide ecological, geographical, and genetic diversity, natural isolates of the budding yeast Saccharomyces have become an important model system for population/evolutionary genomics (Liti, 2015) and to study the complex genetic architecture of lifespan (Kaya et al., 2015; Stumpferl et al., 2012; Kwan et al., 2013; Janssens and Veenhoff, 2016; Jung et al., 2018; Barré et al., 2020). Therefore, these strains offer a powerful genetic pool to understand how natural genetic variation may shape lifespan variation. Accordingly, we assayed the RLS of ~3,000 individual cells representing these isolates under two different conditions: yeast peptone dextrose (YPD, with 2 % glucose), and yeast peptone glycerol (YPG, 3 % glycerol as a respiratory carbon source), and identified up to 10-fold variation in median RLS under each condition. Although little is known about the life histories of these wild isolates they face different, niche-specific evolutionary pressures for adaptation to different stresses (Liti et al., 2009; Liti et al., 2006; Tirosh et al., 2009; Liti, 2015). To understand how different genotypes arrive at different lifespan phenotypes, we further analyzed endophenotypes (gene expression, metabolite abundance) to characterize the molecular patterns associated with condition-specific lifespan variation. Following characterization of transcripts and metabolites with significant association with longevity, we identified pathways associated with median RLS across these isolates. Our data showed that the naturally arising variation in genotype can cause large differences in lifespan, which are associated with distinct patterns of gene expression and metabolite abundances. These analyses revealed connected pathways that have not been previously associated with lifespan variation in a laboratory setting. Altogether, we present the most comprehensive analysis to date of how the environment and genetic variation interact to shape aging and the associated life-history traits. Results Growth characteristics and lifespan variation across wild isolates First, we monitored growth characteristics of natural yeast isolates on standard glucose media (i.e., during fermentation) and on media with a nonfermentable substrate, glycerol, as a carbon source (i.e., during respiration), using an automated growth analyzer, and calculated the doubling time. Most wild isolates grew faster than the diploid laboratory wild-type (WT) BY4743 strain under both conditions, with an average doubling time of 65 min in YPD and 125 min in YPG (Figure 1A, Supplementary file 1). Importantly, most of these strains grew at a similar rate on YPD, and variation in growth rate on YPG was also relatively small among them, indicating that these laboratory-optimized culture conditions are suitable for supporting nutritional needs of these strains and for RLS analysis (Figure 1A). Figure 1 with 1 supplement see all Download asset Open asset Doubling time and replicative lifespan of yeast wild isolates. (A) Distribution of mean doubling time (DT, in minutes) on yeast peptone dextrose (YPD; 2 % glucose) and yeast peptone glycerol (YPG; 3 % glycerol). (B) Examples of lifespan curves for the selected strains. Black curve shows lifespan under YPD conditions, and red curve under YPG conditions. (C) Median replicative lifespan (RLS) distribution across S. cerevisiae (red) and S. paradoxus (turquoise) isolates grown in YPD. (D) Distribution of median RLS across different conditions. Source data are provided as Supplementary file 1. Next, we assayed RLS of these isolates at 30 °C on both growth conditions (YPD and YPG media). Under the YPD condition, we observed a remarkable ~10-fold variation in median and maximum RLS (Pearson correlation coefficient = 0.65 between median and maximum RLS, p < 0.00001) across these isolates (Figure 1B, C, Figure 1—figure supplement 1A, Supplementary file 1). The average median RLS of S. paradoxus strains (29.7) was significantly higher (p = 2.7 × 10−9) than the average median RLS of S. cerevisiae strains (22.7) (Figure 1C, Figure 1—figure supplement 1A). Among the 76 strains analyzed, S. paradoxus strain Y7 showed the longest median RLS (RLS = 42), whereas S. cerevisiae strain Y10 exhibited the shortest median RLS, with 50 % of Y10 cells ceasing division after producing only four daughters (Supplementary file 1). Glycerol as a growth substrate can extend both RLS and chronological lifespan (CLS; Burtner et al., 2009; Kaeberlein et al., 2005a). In the case of CLS, the increased longevity is caused by a switch from fermentation to respiration Burtner et al., 2009; however, mechanisms by which glycerol affects RLS are unclear, since respiratory metabolism is not always required for RLS extension in laboratory WT strains (Kaeberlein et al., 2005a). While we observed a significant (p < 0.05, Wilcoxon rank sum test) median RLS increase in 32 strains on YPG with 24 strains significantly decreased median RLS; and the remaining 7 strains showed no significant changes (Figure 1—figure supplement 1B, Supplementary file 1). For example, S. cerevisiae strain BC187 showed a significantly increased median RLS on glycerol (37.5 in YPG versus 32 in YPD), and S. paradoxus strain KPN3829 also showed a similar increase (21 in YPD and 35 in YPG). On the other hand, S. cerevisiae strain YJM975 showed a significant decrease (median RLS = 31 in YPD and 23 in YPG) (Supplementary file 1). We further dissected the effect of carbon source on RLS by comparing median RLS variation between YPD and YPG conditions. Interestingly, strains with the shortest RLS on YPD tended to achieve the longest lifespans on YPG, while the long-lived strains on YPD generally did not show a further RLS increase (Figure 1—figure supplement 1B, C) (Pearson correlation coefficient = −0.51 between median RLS on YPD and median RLS on YPG, p < 0.0001). A similar observation was shown under CR conditions in the case of single-gene deletions, where the shortest-lived strains tended to yield the largest lifespan extension when subjected to CR (Schleit et al., 2013). Overall, we observed significant differences in lifespan across these strains on both conditions. The observed lifespan pattern on YPG conditions in comparison to those on YPD suggests that long-lived strains might reside in environments with low fermentable carbon sources, so that they undergo distinct metabolic regulation to metabolize respiratory carbon sources. As such, they did not show further RLS extension when grown in glycerol. Endophenotype variation across wild isolates While many previous large-scale omics studies on aging have focused on genome-wide association (Stumpferl et al., 2012; Burke et al., 2014; Wilson et al., 2020; Deelen et al., 2019), recent comparative studies on transcriptomics (Fushan et al., 2015; Fuentealba et al., 2021), proteomics (Yang et al., 2008; Tanaka et al., 2018; Heinze et al., 2018), metabolomics (Laye et al., 2015; Ma et al., 2015b; Cheng et al., 2015), and ionomics Ma et al., 2015a have begun to shed light on molecular patterns and mechanisms associated with the aging process. For example, it has been suggested that natural variation is associated with extensive changes in gene expression, translation, and metabolic regulation, which in turn may affect fitness under different stress conditions (Whitehead, 2006; Vu et al., 2015). In fact, gene expression variation has repeatedly been postulated to play a major role in adaptive evolution and phenotypic plasticity (Vu et al., 2015; Gilad et al., 2006), as well as specific phenotypic outcomes such as changes in morphology (Beldade et al., 2002) and lifespan (Whitaker et al., 2014). Similarly, comparative studies of metabolite profiles have been utilized to describe the genotype to phenotype relations in model organisms (Guijas et al., 2018; Harrison et al., 2020). Accordingly, we aimed to explain lifespan differences among these wild-derived yeast isolates by analyzing their gene expression variation (based on transcriptomics analyses) and differences in their metabolite levels (based on metabolomics analyses). In the case of the transcriptome, we obtained ≥5 million 150 bp paired-end RNA-seq reads for each strain grown on YPD. For metabolomics analyses, we applied targeted metabolite profiling using liquid chromatography–mass spectrometry (LC–MS). After filtering and quality control, the dataset contained RNA-seq reads for 5376 genes and 166 metabolites identified commonly across all isolates (Supplementary file 2). The expression profiles between strains were similar to one another, with Spearman correlation coefficients of strain pairs ranging between 0.59 and 0.93 (except the pairing involving Q59.1, CBS5829, YPS606, and UFRJ50791, with the range between 0.21 and 0.79) (Figure 2—figure supplements 1 and 2). To determine whether the previously published sequence-based evolutionary relationships (Liti et al., 2009) were reflected in their gene expression variation, we constructed gene expression phylograms using a distance matrix of 1 minus Spearman correlation coefficients and the neighbor-joining method (Brawand et al., 2011). The resulting topology of species-specific trends was largely consistent with their phylogeny with a clear separation between S. cerevisiae and S. paradoxus strains (Figure 2A); however, at the intraspecies level, the topology was not consistent with the phylograms of genomic data. Figure 2 with 3 supplements see all Download asset Open asset Endophenotypic variation across strains. (A) Phylogenetic relationship based on the transcriptome data of 76 strains of 2 species. Principal component analysis (PCA) of (B) transcriptomics and (C) metabolomics. Percent variance explained by each principal component (PC) is shown in parentheses. Pathway enrichment analysis for combined top genes and metabolites contributing to (D) PC1 and (E) PC2. Some of the enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways are shown in each panel. Source data are provided as Supplementary file 3. To visualize endophenotypic variation between these two species and across the strains of the same species, we performed principal component analysis (PCA) on each type of data. PCA of the transcriptome revealed a pattern resembling the phylogenetic relationship, with the first three principal components (PCs explaining ~49 % of total variance in gene expression (Figure 2—figure supplement 3A)). Although some S. paradoxus strains clustered with S. cerevisiae, we observed clear species segregation based on PC2 (except for some outlier strains that were separated by PC1). PCA of metabolomics data revealed a similar structure with the first three PCs explaining ~41 % of total variance in metabolite levels, and PC2 somewhat separating the species (Figure 2—figure supplement 3B). To understand the basis of this segregation pattern, we performed pathway enrichment analysis by combining the 500 top genes (250 with positive weights and 250 with negative weights) and 40 top metabolites (20 with positive weights and 20 with negative weights) contributing to each PC, respectively. This integrative analysis of genes and metabolites (see Materials and methods) contributing to PC1 revealed a distinct set of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, including RNA degradation, mitogen-activated protein kinase (MAPK) signaling pathway, cell cycle, pantothenate and CoA biosynthesis, ribosome biogenesis, and pentose phosphate pathway (Figure 2D, Supplementary file 3). The analysis of genes and metabolites for PC2 revealed the KEGG pathways related to ribosome, autophagy, endocytosis, cell cycle, mRNA surveillance, and nucleotide excision repair (Figure 2E, Supplementary file 3). These results suggest that these processes diverged most significantly across the wild isolates of two species of Saccharomyces genus and may account for their phenotypic diversity, including lifespan. It should be noted, however, that we do not know exactly what each PC represents, unless it perfectly aligns or correlates with some known variables. In addition, either biological (e.g., phylogenetic structure), technical (e.g., data normalization or batch effect), or mixed effects of both may render PCA biased (Uyeda et al., 2015). Relationship between endophenotypes and lifespan To identify endophenotypes (transcripts and metabolites) correlating with lifespan variation across wild isolates, we applied the phylogenetic generalized least squares (PGLS) method to account for phylogenetic relationships among the strains and test for different models of trait evolution (Felsenstein, 1985; Freckleton et al., 2002). Regression was performed between endophenotypic values and median RLS under different models of trait evolution and the best-fit model was then selected based on maximal likelihood. To assess the robustness of these relationships, we repeated the regression after taking out one yeast strain at a time and only those regressions that remained significant were further considered. This ensured the overall relationship did not depend on a particular isolate. With the PGLS approach, we identified 73 transcripts with significant correlation with median RLS (padj ≤ 0.01; 39 with positive correlation and 34 with negative correlation) (Supplementary file 2). Among the top hits with positive correlation were a putative zinc finger protein gene (padj = × gene (padj = (padj = × and protein gene (padj = (Figure Supplementary file 2). The top hits with negative correlation included the genes for kinase (padj = and (padj = of protein gene (padj = and gene (padj = (Figure Supplementary file 2). Next, to assess of our hits were previously in yeast lifespan, we extended our of significant genes to genes by a at = and these with the genes associated with RLS in laboratory WT strain in the 2009). genes from the published with effects on RLS or of laboratory yeast strains and (Supplementary file 5376 genes expression was across the wild isolates, were 39 genes present in both our and 23 of which showed the same of correlation with For example, and genes increase RLS when (Supplementary file and are associated with increased RLS when their levels decrease across wild isolates However, the overall pattern of genes as well as the of correlation did not p It should be that many of the genes in are from single-gene studies and has been no comprehensive studies gene on a genome-wide This a that the genes we identified might not be among the genes from other It is also that the genetic architecture of trait variation in natural populations may differ from that which is from studies of including extensive single-gene studies of lifespan variation in yeast et al., 2015). The of between the genes expression correlates with lifespan variation in wild isolates and genes that affect RLS in studies on laboratory WT background this we then trait variation in wild isolates and strains may at the transcriptome in a that may be at the level of gene expression, or at the level of biological To do this we the gene expression patterns across wild isolates with those of laboratory strains et al., RLS was et al., 2015) previously (Figure supplement Supplementary file 1). We calculated an association of gene expression with different of RLS RLS, median RLS, and maximum across strains. Our analysis revealed significant genes (padj = associated with three types of RLS and more than genes associated with median RLS (Figure Supplementary file 2). To the of wild isolates and we then calculated a correlation matrix of gene expression changes across strains and wild isolates (Figure We no positive correlation between gene expression changes across strains and gene expression changes across natural isolates (Figure Figure 3 Download asset Open asset genes expression correlates with median replicative lifespan (RLS). (A) expression level of and correlates with median (B) of and correlates with median Regression p values can be in Supplementary file which is also the source data file for these Figure with 1 supplement see all Download asset Open asset analysis of yeast and wild isolates. (A) genes (padj < associated with and mean replicative lifespan or and across strains based on transcriptomics data obtained from strains. Genes and associated with RLS and are significantly across different of RLS p < (B) correlation matrix of gene expression effects across single-gene strains and those that we across the wild isolates that are associated with coefficient is calculated using of top significant genes for each of signatures with Spearman phylogenetic regression least (C) enrichment of genes associated with RLS across and natural strains. are based on enrichment Supplementary 1 and 2 are provided as source data for these We then performed functional enrichment (gene set enrichment of genes associated with RLS across and wild isolates to see with RLS may at the level of the biological We that the transcripts associated with RLS in these two populations distinct of biological pathways (Figure Supplementary file 2). For the genes correlating with longevity across the the enriched included cellular to ribosome, translation, cellular and repair (Figure Supplementary file 2). On the other hand, enriched in wild isolates included cycle, and metabolic regulation of and (Figure Supplementary file 2). Overall, our comparative analyses of lifespan-associated gene expression signatures in laboratory-adapted yeast strains versus wild isolates suggest that different genetic might have evolved at level across wild isolates to lifespan. Next, we for metabolites with RLS across wild yeast isolates. The a of regulation of both the transcriptome and and it has been for phenotypic variation that lifespan (Laye et al., 2015; Ma et al., 2015b; Cheng et al., 2015). Among 166 metabolites that we 31 exhibited significant association with median RLS (padj ≤ (Supplementary file 2). Among the top and acid (Figure whereas acid showed negative correlation (Figure supplement 1). Our metabolite also included several related with positive correlation to RLS and which are known to be involved in and et al., 2020; et al., 2014). identified transcripts and metabolites associated with lifespan, we aimed to among to better understand biological of lifespan variation and mechanisms of longevity across the wild isolates. Figure with 1 supplement see all Download asset Open asset metabolites correlating with median replicative lifespan (RLS). (A) chromatography–mass spectrometry of and that and (B) of and that with median Regression p values can be in Supplementary file This file is also provided as a source data file for these signatures of lifespan extension across wild isolates To understand the molecular basis of RLS variation in wild isolates, we applied an integrated pathway approach et al., A combined metabolomics and transcriptomics data analysis revealed a role for metabolic regulation of and amino acid as well as and (Figure supplements For example, our data revealed that correlates with RLS (Figure In addition, we that an in the pathway known as the et al., et al., correlates with median RLS (Figure a of degradation, to an increased in long-lived strains. for metabolic regulation of in long-lived strains from our transcriptome data which the first of et al., 2019), was significantly (padj = in long-lived wild isolates (Figure 3B). These a for the observed in long-lived strains. Similarly, a between our transcriptome and data provided into We observed a negative correlation between and lifespan (i.e., long-lived strains to have (Figure Additionally, our transcriptome data showed negative of two
Source: eLife
Language: English
Subfield: Aging
Open Access status: gold
Estimated APC paid: $2,500
OpenAlex ID: W4251305185
Cites:
Cited by:
Related to: 10